Modern Power Supplies: You know what grinds my gears...?

[GWestphal]

Why is that main power cable so flipping huge still. Is there really no way to get a thinner main cable? Why hasn't it been "SATAfied" making that connector smaller and easier to plug would be nice too. Not to mention having them mounted parallel to the board vs perpendicular would be super. While they are at it how about positioning the connector somewhere that isn't retarded, like a horizontal mount on the top. Why do we still have these gross 4 wire chained molex connectors, why not make a nice, easily route-able power cable instead of having a mess of wires.

Don't even get me started on why we still use full size USB ports....why aren't we using something like microUSB or lightning, perhaps with a locking feature if people are concerned about cables getting yanked out too easily?

How about those PCI-E connectors? Do they really need to be that large? We can't ***e up with a more sleek/space efficient connector interface, really? Same with DIMM slots.

It's past the year 2000!!!

\rant

 

[nickbits]

I hate parallel sata connectors, a parallel power supply would be even worse.

What I hate are USB plugs. I alway seem to put them in the wrong way and even the right way they don't seem to fit half the time.

 

[GWestphal]

I actually like Apple's lighting for that reason, small and reversible and it is more solid than microUSB.

Perhaps you're right about the parallel connectors, but have a 90 degree connector on the end of the cable would achieve the same thing so you could route it close to the board.

 

[KentState]

MY pet peeve is that SATA power cables have caught up with SSD's. For example, I have 3 3.5 SSD cradles in my case. These allow me to stack two SSD on top of each other. The SATA power cables are so bulky that they do not fit without putting a bunch of pressure on the connectors. To get around this I purchased a bunch of IDE to SATA power adapter cables to get the correct spacing. Not sure what the solution is, but some type of backplane like adapter needs to be implemented.

 

[Torn Mind]

The molex connector is VERSATILE; countless adapters can be plugged into a molex connector, and for various products. It is beyond funny how something that merely delivers 12v and 5v power to a component gets this much hate do to it being "unfashionable".

There are 24 wires on the main ATX connector, if you want to compromise wire gauge and/or risk sticking it in backwards, be my guest, don't cry when something melts or a fire starts because some it was overwhelmed having to deal with some tri-Crossfire setup.

 

[NutBucket]

The laws of physics cannot be overcome. The only way to make the mb wiring smaller is to increase the voltage and put a lot more power converters on the mb. In other words, it isn't going to happen.

 

[GWestphal]

If cpu/mobo manufacturers could choose their own power delivery standard, I wonder if they would choose the 24 pin connector. I don't know that they would, I think it's just what they plan their architectures around because that is the standard. If you develop a newer one with fewer wires they will plan around that.

 

[Mark R]

Power could be more effectively delivered at a higher voltage - maybe 20V.

It's historical that we still have 5V and 3.3V supplies on the ATX connector, and indeed for hard drives. Virtually no electronics in a modern system operate directly off these voltages, it's all regulated down from the 12V supply, yet a substantial propoprtion of the 24 pin molex connector is dedicated to these supplies.

A 20V single supply to the motherboard would greatly simplify wiring, and avoid the need for the PSU to produce legacy voltages when they are not required.

 

[GWestphal]

I know transferring higher voltages is more efficient, but don't CPUs use like 1V. With spinning disks on the way out, we don't really need to worry about electric motors in computers except for maybe fans but I can see them being replaced by MEMS jet based cooling . What voltages will we need ? RAM, CPUs, and SSDs all use around 1V. Only GPUs need lots of power now. Can you just voltage divide one 20V input into a few 1V rails and a 12V rail, but how does the heat dissipation work with the voltage dividing?

 

[NutBucket]

No you cannot voltage divide as current consumption fluctuates constantly and voltage supply regulation is a tight specification at those low operating voltages.

 

[NutBucket]

Plus a voltage divider would mean every device would have a different ground reference and would need to have floating IO. That's not going to happen.

 

[CM Phaedrus]

Real problem with a voltage divider is that to get 1W at 1V from a 20V supply, you need to burn 19W in your resistors. It's horribly inefficient.

However, it would be trivial, and beneficial in terms of power efficiency and reliability, to have a single 24V power supply. Every IC component except some 5V chips already have a power supply of some sort on the motherboard, GPU, or whatever. Pulling it down from 24V, instead of 12V/5V/3.3V would be trivial, and improve power supply efficiency. Then put a 24V to 12V power supply on the motherboard and power the fans via 4-pin PWM connectors, and HDD motors via modified SATA cables, and you have the simplest system, which requires only one or two cables of ~12 wires to the motherboard and one cable per GPU, and runs more efficiently than modern systems with fewer failure points. PSUs would get cheaper, and motherboards would rise slightly in price, but overall the cost of building a PC would go down.

 

[GWestphal]

shall we call it the AnandTX standard then

 

[86waterpumper]

perhaps with a locking feature

I agree more cables need this. It is ironic that internal cabling such as sata cables have a lock when they are protected and out of the way. External cables such as usb etc do not. I know this is not the video forum, but my biggest pet peeve is hdmi cables. this is the dumbest design. I have yet to have a hdmi device that would stay properly connected. For instance on my roku, the cable weighs more than the device haha. Every day in order to use it I must re connect it or I get a purple screen because of a janky connection. At least usb usually fits tighter.

 

[aigomorla]

Why is that main power cable so flipping huge still. Is there really no way to get a thinner main cable? Why hasn't it been "SATAfied" making that connector smaller and easier to plug would be nice too. Not to mention having them mounted parallel to the board vs perpendicular would be super. While they are at it how about positioning the connector somewhere that isn't retarded, like a horizontal mount on the top. Why do we still have these gross 4 wire chained molex connectors, why not make a nice, easily route-able power cable instead of having a mess of wires.

Don't even get me started on why we still use full size USB ports....why aren't we using something like microUSB or lightning, perhaps with a locking feature if people are concerned about cables getting yanked out too easily?

How about those PCI-E connectors? Do they really need to be that large? We can't ***e up with a more sleek/space efficient connector interface, really? Same with DIMM slots.

It's past the year 2000!!!

\rant

u do know wires have been the same since the conception of the light bulb.

The wires need to be a set gauge thickness to handle the current raited for the unit.

If the wires were thicker, it would make the connector even more bigger to handle the same current of those wires.
If the wires were thinner, you risk fire hazard if u overload the wire.

The Mainboard 24 pin will always be 24pin.
Infact it used to be only 20 pin... however with modern processors and the advent of things like powered USB, the motherboard required 4 additional pins to turn into a 24 pin.

There are Thin ITX setups which have 2 pin plug like a laptop plug.... doesnt require 24 pin.....

Black Plug on the left lower corner...

I dont expect fullsize desktops tho to shrink in number of pins at all because of the layout in which each of those pins need send power to.

The laws of physics cannot be overcome. The only way to make the mb wiring smaller is to increase the voltage and put a lot more power converters on the mb. In other words, it isn't going to happen.

Exactly what i was trying to say up top!



Its not possible with our materials at the electrical end.

Real problem with a voltage divider is that to get 1W at 1V from a 20V supply, you need to burn 19W in your resistors. It's horribly inefficient.

i cant recall of anything on your device which uses 1W.
The board was setup to split a 3W rail for IC's which use less then 1W.

The PSU has 3 rails... like u stated.. 3.3V / 5V / 12V
The board takes those 3 three rails and divides it amungst what is required.
If we had a universal 24V PSU system like you stated... we would have more waste in heat trying to shrink 24->3.3 vs using 3.3 straight from the PSU.
Also we run into a NASTY Bouncer called HEAT... if u want to do the stepdown at the board, well, ur starting to Prep hell for its fiery fury....

There is nothing wrong with the PC PSU.
Infact the PC PSU is considered one of the cream of crops in PSUs because of the EQ it handles.
The only thing which i would say comes close to the PC PSU would probably be a Audio Amp.

 

[GWestphal]

Right that's what I'm saying. Most things are using ~1V, so why are we providing 3.3, 5, and 12V. If all the manufacturers already put power regulators on their hardware to convert from 3.3 or 5 to 1V or what ever, why not just deliver a 1V or 1.5V line and a 12V or something like that

 

[GWestphal]

Just seems that PSU standards need to be updated since hardware is substantially more efficient than it was 20 years ago when the ATX standard was created. So why not change the voltages to account for that and decrease losses in voltage conversions.

 

[aigomorla]

Right that's what I'm saying. Most things are using ~1V, so why are we providing 3.3, 5, and 12V. If all the manufacturers already put power regulators on their hardware to convert from 3.3 or 5 to 1V or what ever, why not just deliver a 1V or 1.5V line and a 12V or something like that

cuz it would mean the board would be required to do that.

That causes additional heat for the board to mitigate, as well as increases the cost..

It would make more sense to put that feature in a PSU, and u buy according to what u need/want.

 

[CM Phaedrus]

Devices do not "divide power" from the 12V/5V/3.3V

The CPU draws +12V power, and uses a buck converter to efficiently reduce the voltage to ~0.9 - 1.6V

The GPU draws +12V power, and uses a buck converter to efficiently reduce the voltage to ~1.0 - 1.4V

The RAM draws +3.3V power, and uses a buck converter to efficiently reduce the voltage to ~0.8 -1.8V

Are you seeing the pattern?

The only devices that draw power from the PSU without passing it through another power supply are +5V motherboard chips, +5V hard drive chips, and +12V hard drive and fan motors. And these devices are relatively lower power (except HDDs, in quantity).


It would be more efficient to provide a single 24V line (which would reduce losses from voltage drop in the cable, and reduce the number and thickness of wires needed) and use buck regulators on the motherboard to power everything else. A power supply would only need to provide 24V and 12V...

 

[GWestphal]

Would the motors work better or more efficiently if they ran at higher voltages? How many chips are designed for 5 or 12 volts simply because that is the ATX standard vs what might be optimal efficiency?

Why not use even higher voltages if efficiency increases with voltage? Why not supply 1000V DC? Or is that too much voltage to buck down or something with dielectric breakdown/arcing etc?

What is a ball park peak efficiency you could get if you could choose the voltages from the power supply and for all the chips?

 

[NutBucket]

Then whole point of reducing the voltage requirements for ICs is to reduce power consumption and heat production all in the name of faster switching speeds. Switching logic levels from 0 to 1.8 volts can be accomplished much easier than 0 to 5. Plus transistors need to be larger to handle more power. So yeah, increasing operating voltage isn't a viable option.

 

[ExcaliburMM]

I'd rather we all took a minute to bitch about the placement of 4 and 8 pin CPU power connectors and how the wire is NEVER long enough to route behind the board in any reasonably sized case. v_v

That being said, yes, SATA power cables suck, but I see someone recommended using those little short SATA to molex adapters for proper routing and spacing, which is actually not a bad idea. Gotta keep that in mind for my next build.

 

[Insert_Nickname]

I'd rather we all took a minute to bitch about the placement of 4 and 8 pin CPU power connectors and how the wire is NEVER long enough to route behind the board in any reasonably sized case. v_v

You can always use an extender cable if you have that particular problem. Easy, cheap and make routing cables easier.

 

[GWestphal]

ok, that makes sense, what are the limitations of CPUs currently? Why do they use 1.8v versus 10nV?

 

[NutBucket]

There needs to be enough voltage swing to overcome noise for one thing. And the feature size (etching size of the wafer) is still too large to allow ultra-low voltages. So the limitation is physical.

The 1.8V I stated is just an example. I haven't kept current with CPU technology in quite a few years but I do recall core voltages being in the 1.2V range? I'm sure someone will correct me.